See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228346619

Developing And Validating The Females In Mathematics Scale, FIMS

Article · September 2010

CITATIONS READS
0 83

3 authors:

Katherine Picho Jerome Katrichis

26 PUBLICATIONS   264 CITATIONS   
University of Hartford
19 PUBLICATIONS   486 CITATIONS   
SEE PROFILE
SEE PROFILE

D. Betsy Mccoach
University of Connecticut
146 PUBLICATIONS   5,472 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cognitive load theory and the design of education View project

Growth Mixture Modeling View project

All content following this page was uploaded by D. Betsy Mccoach on 31 May 2014.

The user has requested enhancement of the downloaded file.

 77
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Developing and Validating the Females In Mathematics Scale (FIMS)

Katherine Picho
University of Connecticut, USA

Jerome M. Katrichis
University of Hartford, USA

D. Betsy McCoach
University of Connecticut, USA

Abstract

Over the years, research on teacher bias towards females in mathematics has yielded
different results, which the study attributes to the narrow definition and measurement of
bias. This study validated the Females In Mathematics Scale (FIMS), to measure teacher
bias against females in mathematics. FIMS, based on the tripartite model of attitudes and
constructed using within-method triangulation, demonstrated acceptable levels of internal
reliability, and discriminant validity. Preliminary analyses showed that despite considerable
variability in what teachers think and how they feel about female students in mathematics,
they are more likely than not, to engage in unbiased behaviors in classroom interactions.

Keywords: Exploratory factor analysis, teacher attitudes, gender stereotypes, teacher

beliefs, mathematics, tripartite model

The paucity of instruments with strong psychometric properties in

educational research has raised concerns from academics about both the rigor of
methods used and the validity of findings resulting from these methods for quite
some time. Research on teacher attitudes, specifically bias towards students
belonging to marginalized groups such as females in mathematics, and students with
disabilities has been a perennial subject of discussion mainly because of mixed
results yielded over several years of research. Some researchers provide evidence
for the existence of teacher bias and its harmful effects on student performance
(Catsambis, 1995; Inzlicht & Ben-Zeev, 2003; Sadker & Sadker, 1990), others
acknowledge the presence of such bias but argue that expectancy effects resulting
from bias are relatively small, accounting for 5-10% of the variance in student
achievement (Jussim, 1991; Jussim & Eccles, 1992) and still, a few other studies
have found no evidence to suggest that bias exists (Altermatt, Jovanovic, & Perry,
1998; Wheatley & Jones, 1990). The disparity in findings reflects potential
problems in the validity and/or reliability of the measurements used, possibly due
to a narrow conceptualization of attitude in the design of instruments used in the
studies, and hence the limited use of triangulation in measurement.
The majority of studies on teacher bias have used dichotomous measures
(Rosenfeld & Rosenfeld, 2007; Tiedman, 2000) to ascertain its existence even
though research delineates attitude as a construct existing in gradations (Albarracin,
Johnson & Zanna, 2005; Brophy & Good, 1974). The examination of bias through

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 78
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

a categorical lens has a significant impact on the way bias is measured because it
distills the phenomenon to its most basic form (where it is either present or absent),
and as a result restricts the capabilities of what instruments can measure. We
believe that „mixed results‟ generated over years of research are largely a function of
narrow definition and measurement of bias, and we contend that the development
of robust instruments capable of capturing bias holistically hinges on how
adequately bias is operationalized and how well its nature as a continuous variable is
taken into account in instrument design or research methodology.
According to the Brophy and Good model (1974), teacher attitudes occur
along a continuum from „proactive‟ to „over reactive‟ with „proactive‟ teachers most
inclined to being unbiased and vice versa. Good (1987) contends that rather than
exist primarily at these extremes, teachers possess attitudes which vary in degrees
along the continuum, the majority of who hold light expectations which are
constantly adjusted as new information about the student emerges. Nevertheless,
subsequent research (Altermatt, Jovanovic and Perry, 1998; Cornbleth & Korth,
1980; Hartley, 1982; Tiedman, 2000) has focused mainly on investigating the
presence or absence of bias. To the best of our knowledge there have not been any
studies that have taken into account the varying degrees of its existence in its
measurement. Findings on bias should therefore be construed, not as „mixed‟ or
inconsistent, but as a testament to the nature of bias as a continuous variable and
the limitations of current instruments to capture this characteristic. So far results on
bias from the aforementioned studies point to individuals who lie at either end of
the spectrum, and in both cases, measures have not been designed to account for
the larger part of the story--one arguably central in reconciling both perspectives--
the mid-section where the majority of teachers lie. Incorporating this missing piece
of the puzzle in measurement requires that the operational definition of bias be
expanded to include all its dimensions in measurement (i.e. its cognitive, affect and
behavioral components).
Although the tripartite model defines attitudes as comprising cognitive,
affective and behavioral components (Olson & Zanna, 1993), most studies have
used one (Catsambis, 1995; Inzlicht & Ben-Zeev, 2003; Jones & Wheatley, 1990;
Sadker & Sadker, 1990) or sometimes two (Evertson, Brophy & Good, 1973; Good
& Brophy, 1972; Silberman, 1969), but not all three components in their research.
A multi-dimensional approach to the operationalization and measurement of bias
brings us closer to providing a more wholesome picture of the phenomenon. It also
provides an avenue to use within method methodological triangulation to
strengthen the validity of research findings. Triangulation uses multiple methods
and measurement procedures in research in order to increase validity (Ma &
Norwich, 2007). Methodological triangulation can be conducted using between
methods or within methods techniques where the former involves using contrasting
research methods to investigate a phenomenon e.g. using both a questionnaire and
observation in a single study (Denzin, 1970). In within methods triangulation,
varieties of the same method are used to investigate a research problem such as self-
report questionnaires with two contrasting scales to measure a construct (Jick,
1979). By providing cognitive, affective and behavioral bases upon which attitudes
can be evaluated; the tripartite model of attitudes enables the use of within methods

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 79
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

triangulation to not only investigate teacher bias along these components but also
strengthen the validity of these findings.
Here we briefly analyze methodological issues involved in the previous
measurement of teacher bias and attempt to remedy these issues by developing and
validating an instrument to measure teacher bias towards females in mathematics.
The validation of the Females in Mathematics Scale, FIMS, uses the tripartite
model of attitudes to expand the measured dimensions of bias, and within-method
triangulation to increase validity. Teacher bias is not only limited to females in math
and the sciences but also extends to males in subjects that are deemed
stereotypically „female‟ like reading (Hartley, 1982; Palardy, 1969), and to children
from lower socio-economic backgrounds (Alexander, Entwisle & Dauber, 1993;
Alvidrez & Weinstein, 1998). However, we chose to develop the Females in
Mathematics Scale (FIMS), because of the abundance of already existing literature
in the field on this topic, reports that teacher bias has a particularly profound
impact on performance and learning outcomes of girls in stereotypically male sex-
typed school subjects, such as math and science (Trouilloud et al, 2006), the
chronic scarcity of females pursuing careers in these domains (National Science
Foundation, 2000), and the fact that mathematics remains a critical component and
determinant of entry into careers in physical science domains like engineering.
Because teacher beliefs are filters that drive decisions regarding instruction (Fang,
1996) and have a significant impact on learning goals and outcomes of all students
(Eccles, 1993; Trouilloud, Sarrazin, Martinek, & Guillet, 2002), it is important that
it be correctly identified and reliably measured in order to design appropriate
interventions or enhance existing professional development programs.
Measurement and interventions could occur at any one of a number of levels from
state or district, to school or grade level. Additionally, once these interventions are
applied, a reliable, psychometrically sound instrument will also allow for the
measurement of attitude change and evaluating the effectiveness of such programs.

Measuring Teacher Attitudes

Teacher bias has been attributed to low teacher expectations of students

belonging to stigmatized groups, and the underperformance of these students has
been linked to teacher behavior stemming from their expectations of these students
(Brophy, 1987). These biases are based, not only on preconceived notions about
student abilities but also based upon how teachers feel towards students; e.g.,
students that teachers are attached to and concerned about receive more teacher
praise, less criticism and higher quality process questions than other students (Good
& Brophy, 1972; Silberman, 1969). Studies on teacher beliefs have been conducted
either through observation (Good & Brophy, 1972; Altermatt et al., 1998) or by the
use of self-reports (Cook, Cameron & Tankersley, 2007; Cornbleth & Korth, 1980;
Rosenfeld & Rosenfeld, 2008). In both methods comparisons between contextually
stereotype-relevant groups (minorities, females, students with disabilities) and other
groups have been central in determining the existence or non-existence of bias
(Cornbleth & Korth, 1980; Hillman & Davenport, 1978; Jones & Wheatley, 1990).
Observational studies have dwelled on differences in teacher behavior in teacher-

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 80
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

student interactions (Altermatt et al, 1998; Evertson, Brophy & Good, 1973; Good
& Brophy, 1972; Silberman, 1969) while self-report studies have relied mainly on
comparisons of teacher ratings of students‟ ability or achievement potential to
report teacher bias (Tiedmann, 2000; Alexander et al, 1993; Alvidrez & Weinstein,
1998).

Measures from Observational Studies. The results from observational

studies have been mixed. Some studies indicated that teachers not only
overestimated the ability of boys (Jussim & Eccles, 1992), but they also spent more
quality academic time with them (Catsambis, 1995; Inzlicht & Ben-Zeev, 2003;
Sadker & Sadker, 1990). Other studies found no evidence of gender bias in teacher
judgments ((Hoge & Butcher, 1983; Hoge & Coladarci, 1989) or teacher-student
interactions among students in the classroom (Altermatt, Jovanovic and Perry,
1998).
Altermatt et al., (1998) observed six science teachers (3 men, 3 women)
across six classrooms and 165 students for 40 minutes each class over one year to
investigate gender differences in both the nature of questions asked (high ordered
versus low-ordered) and the frequency with which either gender was asked to
respond to questions. After factoring in volunteering rates, the authors found no
difference in response rates of the students who responded the most; 53% of total
male volunteering compared to 48% for females, and no differences in the nature
of questions asked. These findings are corroborated by similar studies, which have
found no significant sex differences in opportunities students had to answer abstract
questions (Hillman & Davenport, 1978; Jones &Wheatley, 1990). Results based on
observational studies seem to vary, possibly because considerable subjectivity is
involved in the interpretation of behavior. Second, behavior is highly contextual
and there could be other factors influencing teacher behavior besides bias.
Altermatt et al. (1998) proposed an alternate theory: teacher behavior may be
responsive to and influenced by student behavior within the classroom, like gender
differences in student volunteering rates rather than bias itself. Additionally,
although observational studies provide rich accounts of the phenomenon being
studied, the likelihood that the behaviors observed during these periods do not
closely mimic routine behavior cannot be ruled out with certainty. This then raises
concerns about ecological validity. For these reasons, considerable caution should
be exercised in making generalizations about teacher bias based only on
observation of teacher behavior. A more desirable alternative would be to use
methodological triangulation, where observation in tandem with other methods is
utilized to investigate and strengthen the validity of findings on bias.

Self-report Measures. Because research on bias has focused largely on

teacher-student interactions, the use of self-reports has not been as common in
research on teacher attitudes as observational studies have been. Nevertheless, self-
report data have been used to explore teacher beliefs about „weak‟ students
(Rosenfeld & Rosenfeld, 2007), females in mathematics (Tiedmann, 2000), racial
minorities (Cornbleth & Korth, 1980) and students with disabilities (Cook,
Cameron & Tankersley, 2007). While the reliability of findings based solely on

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 81
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

observational methods have been somewhat tainted by the subjectivity involved in

interpreting results, the reliability of self- reports are rooted in the psychometric
properties of the instruments used. Methodologists recommend that instruments be
validated prior to their application in research studies (Henson, 2006) for good
reason: the rigorous validation process involves item modification guided by the use
of robust factor analytic techniques and validation with several hundred teachers.
When conducted correctly, instrument validation allows researchers to develop
reliable instruments with strong psychometric properties that enhance overall
reliability and validity of findings. Without validated instruments, inferences drawn
about teacher bias are tenuous, especially with small sample sizes. Unfortunately, a
good number of studies have not validated instruments or failed to report sufficient
information about psychometric properties of their instruments, making it difficult
to evaluate reliability and validity evidence. With the exception of Cook, Cameron
and Tankersley‟s (2007) instrument rating teacher‟s attitudes towards students with
disabilities, the measures and reports on teacher bias have been based on non-
validated instruments (Cornbleth and Korth, 1980; Rosenfeld & Rosenfeld, 2007;
Tiedmann, 2000).
Cornbleth and Korth (1980) found evidence of teacher bias towards black
and female students in integrated classrooms. However, detailed information on the
development of the items or the theoretical basis for the attitudinal categories used
to develop the 12-item instrument was not provided, making it difficult to evaluate
content validity. Additionally, inferences about bias were based on evaluations of a
sample of six student teachers, which is so small that any conclusions drawn from
the study are likely to be non-generalizable. Similarly, Tiedmann‟s (2000) measure
of teacher beliefs as predictors of children‟s concept of their mathematical ability in
elementary school sampled only 28 teachers across 28 elementary school classes
with a one-item measure asking teachers to evaluate students‟ mathematical ability
on a 5-point scale. Here single responses were matched to actual math grades for
189 students and used to report higher ability perceptions for boys than for girls
despite gender similarities in performance. However, this study used a small sample
size in conjunction with single-item measures; this combination tends to yield non-
replicable results (Fabrigar, Wegener, MacCallum & Strahan, 1999). Hence it is
possible that methodological artifacts influenced the findings in these studies.
When it comes to instrument validation using factor analytic techniques,
researchers suggest having at least 3-5 measured variables representing each
common factor included in a study (MacCallum, 1999; Velicer & Flava, 1998) and
a sample size of at least 200 for items with communalities less than .6 (MacCallum,
Widaman, Zhang & Hong, 1999) or a ratio of 5-20 respondents per variable
(Stevens, 1996). They recommend and the use of principal axis factoring in the first
stages of instrument development because it identifies latent constructs underlying
measured variables- a primary goal at this stage. A look at the literature finds little
evidence of such practices: Fabrigar et al. (1999) reviewed articles published from
1991 through 1995 in two prominent journals known for methodological rigor with
surprising results. About 40% of analyses did not include reports of the reliability, a
large number of which were single-item measures. In numerous cases, the sample
sizes used were so insufficient that they greatly increased the likelihood of obtaining

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 82
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

under determined factors. They also found that approximately half of the published
applications reported used the wrong factor analytic technique (principal
components analysis vs. principal axis factoring) given their research goals. The
study by Cook and colleagues (2007) faced similar challenges; they used a sample
of 50 teachers to validate a four- item, four-factor scale (one item per factor)
measuring teachers‟ attitudes towards students with disabilities. Teachers rated a
total of 156 students with disabilities and 4 students without disabilities in each of
their classrooms on a 4-point Likert scale with responses ranging from not at all
true to extremely true for four statements: “I would like to keep this student for
another year for the sheer joy of it”; “I would like to devote all my attention to this
student because he or she concerns me”; “I would not be prepared to talk about
this student if his or her parents dropped by for a conference”, and lastly “If my
class was to be reduced, I would be relieved to have this student removed.” The
Pearson correlations for attachment, concern, indifference and rejection ratings are
quite good (values of .77, .70, .71 and .74 respectively) and compared to existing
measures, a considerable level of rigor was involved in the development of this
instrument. However, the authors used single item measures for each of the factors
and didn‟t use any of the conventional tools suggested for factor analysis to
determine factor structure of the items. As such, we have no way of knowing if their
items reflect the constructs being measured.
It appears that the development process of instruments to measure teacher
attitudes have been less rigorous than desired. Apparently measures of teacher bias
have been plagued with more serious problems than previously assumed:
unsatisfactory sampling procedures, relatively lax methodological procedure,
instruments with either unreported or weak psychometric properties, and restricted
range in measurement due to partial definition of bias. Earlier on, the use of
triangulation in measurement and research to improve reliability and validity of
findings was advocated, and the studies reviewed above make a clear case for that.
However, given the aforementioned problems instruments in research on bias
possess, it is unlikely that the use of triangulation with existing measures would
improve and add to the analysis. To do so would approximate pouring new wine
into old wineskins. Addressing teacher bias and remedying issues related to its
measurement would require that researchers return to the drawing board and
develop measures that: look at bias as a multi-dimensional construct, take into
account degree with which bias exists and extend its measurement beyond
dichotomy, follow stringent methodological procedure in research design, and
finally, subject instruments to rigorous validation to determine and strengthen
psychometric properties before use in research. This task, while daunting, is by no
means impossible and certainly carries with it rewards beneficial to strengthening
the overall quality of findings in research on teacher attitudes.

Developing the Females In Mathematics Scale (FIMS)

Attitude research advocates the integration of cognition in the measurement

and evaluation of attitudes (Antonak & Livneh, 2000). The tripartite model of
attitudes, which incorporates all three components of attitude (cognition, affect and

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 83
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

behavior), was used as a foundation for the creation of FIMS. The cognitive
component of attitude was defined as an individual‟s ideas, thoughts, perceptions,
beliefs, opinions or mental conceptualizations of the referent (Findler, Vilchinsky &
Werner, 2007); affect as the amount of positive or negative feelings one has towards
the referent, and behavior as one‟s intent or willingness to behave in a certain
manner toward the referent, or the actual behavioral response (Cook, 1992).

Method

The primary goal was to create items that would minimize response bias
and capture variability in responses along the seven point Likert scale, which ranged
from „strongly disagree‟ to „strongly agree‟. Items corresponding to each of these
three components of attitudes, specific to teaching were created and operationalized
as follows:

(1) Cognition: What teachers think about females‟ interest, effort and ability
in mathematics. High values on the seven- point scale for this factor would indicate
unbiased beliefs and a positive attitude about females in math and vice versa.
(2) Affect: Teachers‟ feelings about females‟ level of engagement, ability and
effort in their mathematics classes. Emotions from the pleasant-unpleasant axis of
the circumplex model of emotions (Russell & Barrett, 1999) were used to generate
items for affect. Teachers rated their feelings about negative levels of student
engagement, effort and performance in their mathematics classes. High scores
denote teachers‟ strong unpleasant feelings about poor performance or low
engagement and effort levels of females in mathematics, indicative of high
expectations and hence a more favorable attitude and vice versa.
(3) Behavior: Items that loaded under this factor concerned how teachers
respond to females in their mathematics classes. Items for this subscale were based
on Good‟s (1987) work regarding how teachers communicate expectations to
students and also on the Good and Brophy (1974) model. Items evaluated
differential teacher behavior in interactions between perceived high and low
achievers such that high values indicated biased behaviors and vice versa.
To summarize, unbiased teachers would have very high scores on cognition and
affect, and low scores on behavior.
Fifty items were created and used in a pilot study with 100 pre-service and
practicing math teachers. These items made direct comparisons between males and
females such as, “Females have a harder time focusing on math than males do.”
Participants also provided qualitative feedback on the items at the end of the
survey. Both qualitative feedback and results from the factor analysis were used in
tandem to assess the strength and uni-dimensionality of the items. Results signaled
weaknesses in the way items were worded.
Although the factor analysis yielded factors that denoted cognition, affect
and behavior, an extra factor emerged that contained a mix of cognitive and
behavior items. However, the cognition items had to do with „behavioral‟ indicators
of females in math e.g. Females don‟t put as much effort in math as males and
females have harder time focusing on math than males. These comparison

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 84
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

statements, even though intended to convey a single meaning, appeared to have

been interpreted multiple ways by different participants, subsequently yielding
numerous multidimensional items in the factor analysis, with approximately equal
loadings on one or two other factors. For example, selecting „strongly disagree‟ for a
statement like „Females are more engaged in math than males‟ was interpreted in
one of two ways: that either (a) females were not as engaged as males were (our
intended meaning) or that (b) both males and females were equally engaged in
math. Therefore, the meaning of both disagree and neutral were unclear. There
was a general lack of variability in responses to the items. That is, individuals
tended to either strongly disagree or strongly agree with items, with very few
selecting more moderate response choices. Participants also indicated confusion at
the implications of selecting „neutral‟ for items like the aforementioned. Selecting
„neutral‟ for „females are more engaged in math than males‟, for example, was
construed in myriad ways by participants: (a) that they did not know if females were
more engaged than males, (b) that they were not sure whether females were more
engaged than males, or that (c) males and females were equally engaged in math. In
sum, using direct comparisons between males and females led to items that were
unclear, subsequently restricting the range of responses. For example, all the affect
stems referenced gender differences but these questions did not account for
teachers who might not perceive any gender differences in or out of their
classrooms. To eliminate this ambiguity and mitigate response bias, an entirely new
set of items that referenced only females was generated for this study.

Content Validity

There were 33 items (sixteen cognition, seven behavior and ten affect) went
through three rounds of content reviews: a peer review, content validation by five
experts, and a final review by the investigators. The content experts checked either
„not relevant‟, „somewhat relevant‟, or „very relevant‟ to gauge the relevance of each
item as a measure for cognition, affect or behavior. A decision had been made a
priori to make 75% the appropriate cut off for correct item placement, with the
exception only for items that were believed to contribute conceptually to the
factors. Any items placed in the correct category by less than 75% of the experts
were deleted from the final instrument and Content Validity Indices, CVI‟s
(McKenzie, Wood, Kotecki, Clark & Brey, 1999) were computed based only on
the percentages in the „very relevant‟ category. But for one affect item, all items had
CVI‟s of 0.8 and above. Cognition and Behavior generated equally high scores
while „affect‟ had the lowest scores. Based on qualitative feedback from the experts
and scores on the CVI, three negatively worded items in the hypothesized „affect‟
factor were deleted, and replaced with positively worded items. The final number
of items remained the same.

Sample and Procedures

The final survey consisted of thirty-three items: 16 dealing cognition, 7 for

behavior, and 10 for affect. Items were rated on a seven point Likert scale; with

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 85
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

responses ranging from strongly disagree to strongly agree. A convenience sample

of 195 mathematics teachers teaching at the elementary to high school level, in
school districts in Connecticut were contacted via email by the State Department of
Education and requested to complete the FIMS online. Responses were received
from a total of 195 mathematics teachers in 35 school districts representing a mix of
urban (30.7%), suburban (54%) and rural schools (15.3%) of low (26.9%), middle
(56%) and high (17.1%) economic status. The majority of respondents (97.2%) were
from public schools. 58.2% of the sample was from High school, 24.3% from
Middle school and 17.5% were from elementary School. A little more than two
thirds of the sample was female (67.9%). The racial composition of the sample,
which is representative of teachers in Connecticut, was predominantly White
(92.7%). Other ethnicities represented included Asians (2.6%), African Americans
(2.1%), Hispanic (0.5%) and 2.1% listed their ethnicity as „other‟.
An exploratory factor analysis using principal axis factoring with oblique
rotation was conducted. Principal axis factoring (PAF) was used instead of principal
components analysis (PCA) because the latter is a data reduction technique that
does not discriminate between unique and shared variance-- and as a result latent
factors aren‟t the focus of the analysis (Henson, 2006). PAF was also used because
it is less likely than PCA to inflate estimates of variance accounted for by the given
factors (Henson, 2006). Also, the use of oblique rotation is recommended if
correlations between factors are expected (Pett, Lackey & Sullivan, 2003), as was
the case with factors in the FIMS.
Five different procedures under the PAF were used to determine the factor
structure: Kaiser-Meyer-Olkin measure of sampling adequacy, eigenvalues greater
than one, scree plot, parallel analysis (PA) and the minimum average partial (MAP)
test. PA and the MAP test are not only the more accurate methods of factor
extraction, but also complementary to each other: in the worst-case scenario parallel
analysis tends to over extract while the MAP test tends to do just the opposite
(Hayton, Allen & Scarpello, 2004). Hence the use of both methods in guiding
decisions about factor extraction reduces the risk of over or under extraction.

Results

The FIMS KMO value of 0.863 suggested it was reasonable to run a factor
analysis on the data and Bartlett‟s test of sphericity was statistically significant (p
<.001). The eigenvalues above one reported seven factors and the total variance
explained by all seven factors was 52.91%. The scree plot (see Fig.1), parallel
analysis and MAP test, however, all suggested extracting three factors.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 86
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Figure 1
Scree plot

Scree Plot

10

8
Eigenvalue

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Factor Number

Three Factor Solution. A three-factor principal axis factor analysis with

oblimin rotation was performed and the solution converged in 16 iterations,
explaining 43.25% of the total variance. Factor 1 (cognition) explained 27.5%, factor
2 (behavior) explained 11.49% of the variance and factor 3 (affect) explained 4.71%
of the total variance.

Analysis. Initially there were sixteen cognition, seven behavior and ten affect
items. After the EFA was run, the pattern matrix showed that 17 items loaded on
cognition, 10 on behavior and 5 on affect suggesting potential problems with cross
loadings among factors. In addition to this, a close examination of the factors in the
pattern matrix shown in table 1 revealed evidence of multi-dimensionality among
factors, discussed in detail below.

Table 1
Pattern Matrix for the Three-Factor Solution

Cognition Behavior Affect

q39. Most females are motivated in math .88
q30. Most females are engaged in math class .81
q22. Most females like taking math classes .81
q41. I am pleased by the level of effort most females devote to
.80
math
q34. Most females spend the time necessary to master
.76
mathematics
q28. Most females work hard at math .74
q37. I am delighted with the performance of most females in
.73
math

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 87
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Cont. Table 1
.71
q11. Most females find math enjoyable.
q14. Most females exert the effort required to succeed in
math .71

q20. Most females do well in math .70

q27. I am happy with the level of interest most females show in
.69
math
q15. Most females are interested in math .69
q10. Most females are focused in math classes. .66
q40. Most females are capable of mastering advanced math
.53
topics with ease
q13. Understanding math concepts comes naturally for
.53
females
q9. In general most females have an easy time learning math
.50
concepts
q26. Females who excel in math are rare .37
q23. I tend to provide additional step by step support to
.74
females when solving simple math problems
q38. I give females additional support as they work through
.72
math problems
q16. I give females extra time to solve math problems .58
q31. When females struggle with math problems, I'm quick to
.56
provide them with clues to the solution
q19. I feel sorry for females who do not do well in math .53
q18. When correcting females in math, I try not to sound too
.43
negative
q29. I pity females who do not succeed in math .41
q25. I encourage females to do more math practice outside
.37
class
q33. I regret my efforts when females show no interest in math .34
q24. Usually, when a girl does poorly in math, it's because she
does not have the ability
q17. I am disappointed when females are bored in math class .35 .63
q32. It is upsetting when females neglect math .31 .58
q35. I get frustrated when my attempts to get females involved
.33 .55
in math do not work
q21. I take it personally when females in my class do not meet
.41 .44
achievement standards in math
q36. Usually, when a girl does poorly in math, it's because she
.41
hasn‟t put forth adequate effort
q12. I'm very direct when providing negative feedback to
females
Note. Extraction method: Principal axis factoring. Rotation method: Oblimin with Kaiser
normalization. Rotation converged in 16 iterations. Suppressed values below .30

Cognition. All but two of the original cognition items loaded under the
hypothesized factor. Of the two items that failed to load on this factor, one loaded
under affect (q36) and the other (q24) did not load on any factors at all. Also, three
affect items cross- loaded very highly on cognition (q27 =.76, q37 =.73 and q41 =
.80). A possible explanation for this is that the wording tapped into both constructs
inadvertently contained a cognitive component. That is, they seemed to evaluate

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 88
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

one‟s thoughts about one‟s affect also (e.g. I am delighted by the performance of
females in math) toward females in math.

Behavior. Similarly, the behavior items appeared to be quite cohesive; but

one of the original items loaded under the hypothesized behavior factor. The item
that did not load on this factor (q12) also did not load on any other factors.
However, seven affect items cross-loaded onto behavior (q17,q19, q21, q29. q32,
q33, q35), and just as in the cognition factor, these items seemed to have wording
problems. That is, the items were worded in such a way that they contained
behavioral components (e.g. I feel sorry for females who do not do well in math).

Affect. The affect subscale performed less optimally than cognition and
behavior subscales. Only four of ten items originally created for the scale loaded on
the factor, with seven others as already discussed, cross loading on cognition and
behavior. The items under affect were also multidimensional loading on both
affect and behavior (see table 2).
The cognition factor emerged strongest and explained the most variance
(27.5%) in teacher beliefs. Affect items performed relatively poorly compared to
the rest of the other items because of the way they were worded. Affect was
assessed relative to teachers‟ behaviors or thoughts about females in mathematics
and this seems to have been the underlying cause of multi-dimensionality or cross
loading in this case. Nevertheless, after excluding the cross loaded items, the
pattern and structure matrices for the three-factor solution respectively, also showed
moderate to high coefficients between the items and their respective factors.

Table 2
Structure Matrix for the Three-Factor Solution

Cognition Behavior Affect

q39. Most females are motivated in math .89
q30. Most females are engaged in math class .82
q41. I am pleased by the level of effort most females devote
.80
to math
q22. Most females like taking math classes .80
q34. Most females spend the time necessary to master
.76
mathematics concepts
q37. I am delighted with the performance of most females in
.76
math
q28. Most females work hard at math .75
q14. Most females exert the effort required to succeed in
.72
math
q27. I am happy with the level of interest most females show
.71
in math
q11. Most females find math enjoyable. .70
q20. Most females do well in math .68
q15. Most females are interested in math .66
q10. Most females are focused in math classes. .66

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 89
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Cont. Table 2
q13. Understanding math concepts comes naturally for .52
females
q40. Most females are capable of mastering advanced math
.52
topics with ease
q9. In general most females have an easy time learning math
.50
concepts.
q26. Females who excel in math are rare .39
q23. I tend to provide additional step by step support to
.72
females when solving simple math problems
q38. I give females additional support as they work through
.71
math problems
q16. I give females extra time to solve math problems .57
q19. I feel sorry for females who do not do well in math .55
q31. When females struggle with math problems, I'm quick to
.55
provide them with clues to the solution
q25. I encourage females to do more math practice outside
-.31 .44
class
q29. I pity females who do not succeed in math .44
q18. When correcting females in math, I try not to sound too
.43
negative
q33. I regret my efforts when females show no interest in
.39
math
q24. Usually, when a girl does poorly in math, it's because
she does not have the ability
q17. I am disappointed when females are bored in math class .38 .67
q32. It is upsetting when females neglect math .36 .62
q35. I get frustrated when my attempts to get females involved
.38 .60
in math do not work
q21. I take it personally when females in my class do not meet
.47 .50
achievement standards in math
q36. Usually, when a girl does poorly in math, it's because she
.41
hasn‟t put forth adequate effort
q12. I'm very direct when providing negative feedback to
females
Note. Extraction Method: Principal Axis Factoring. Rotation Method: Direct Oblimin with Kaiser
Normalization. Suppressed values below .30

Pattern and Structure Matrices. Item-factor loadings on the pattern matrix

were moderate to strong and ranged from .49 - .88, .40 -.74 and .40 -.60 for
cognition, behavior and affect respectively. The moderate to strong range of values
for these items relative to their factors suggests a strong relationship between the
items and their respective factors after controlling for other un-related factors and
variables. Similarly, the structure matrix shown in table 2 revealed moderate to
strong item-factor loadings for cognition (.50 - .89), behavior (.44 - .72) and affect
(.30 -.65), indicating moderate to strong correlations between items and their
respective factors. There wasn‟t that much difference between the pattern and
structure matrix with respect to the item- factor loadings, which suggests low
correlations between the factors.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 90
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Factor Correlations. Table 3 shows factor correlations between cognition,

affect and behavior. Cognition and behavior have a low, negative correlation
because they were scored in opposite directions. That is, behavior items were
scored such that high values were indicative of bias while for cognition, the reverse
would be true. The relationship between cognition and affect appears to be weak
and negative (implying that those who report more unbiased thoughts about
females in mathematics actually harbor more biased feelings towards these
females). Similarly, the positive and relatively moderate correlation between
behavior and affect would suggest that those who report unbiased feelings about
females in mathematics actually tend to behave in a more biased manner towards
these females. There seems to be a discrepancy regarding the relationship of affect
with the other two factors. This could either be a result of either social desirability
in self-reporting or a methodological artifact due to poor wording of affect items, or
both. The low correlations between the factors as shown in the factor correlation
matrix above are, however, indicative of discriminant validity.
Table 3
Factor Correlation Matrix

Factor Cognition Behavior Affect

Cognition __ -.12 -.06
Behavior __ .15
Affect __

Communalities. Overall, communalities-- the variance in each item

explained by its factor-- were moderate to high (Fabrigar et al., 1999) for cognition
(.29 -.79.), behavior (.21 -.55) and affect (.23 -.68).

Table 4
Communalities for the three- factor solution

Initial Extraction
q9. In general most females have an easy time learning math concepts .52 .29
q10. Most females are focused in math classes. .66 .44
q11. Most females find math enjoyable .71 .54
q12. I'm very direct when providing negative feedback to females .37 .08
q13. Understanding math concepts comes naturally for females .48 .31
q14. Most females exert the effort required to succeed in math .69 .56
q15. Most females are interested in math .66 .51
q16. I give females extra time to solve math problems .46 .37
q17. I am disappointed when females are bored in math class .57 .54
q18. When correcting females in math, I try not to sound too negative .43 .21
q19. I feel sorry for females who do not do well in math .44 .34
q20. Most females do well in math .63 .54
q21. I take it personally when females in my class do not meet achievement
.50 .40
standards in math
q22. Most females like taking math classes .74 .66
q23. I tend to provide additional step by step support to females when solving
.55 .55
simple math problems

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 91
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Cont. Table 4
q24. Usually, when a girl does poorly in math, it's because she does not have
.20 .03
the ability
q25. I encourage females to do more math practice outside class .42 .32
q26. Females who excel in math are rare .49 .33
q27. I am happy with the level of interest most females show in math .66 .55
q28. Most females work hard at math .75 .57
q29. I pity females who do not succeed in math .40 .23
q30. Most females are engaged in math class .72 .68
q31. When females struggle with math problems, I'm quick to provide them
.42 .29
with clues to the solution
q32. It is upsetting when females neglect math .47 .46
q33. I regret my efforts when females show no interest in math .43 .23
q34. Most females spend the time necessary to master mathematics concepts .70 .59
q35. I get frustrated when my attempts to get females involved in math do not
.48 .48
work
q36. Usually, when a girl does poorly in math, it's because she hasn‟t put
.39 .24
forth adequate effort
q37. I am delighted with the performance of most females in math .73 .64
q38. I give females additional support as they work through math problems .54 .51
q39. Most females are motivated in math .80 .79
q40. Most females are capable of mastering advanced math topics with ease .54 .31
q41. I am pleased by the level of effort most females devote to math .78 .68

Reliability analysis. Based on the results from the factor analysis, items to
conduct a reliability analysis were selected. An item was excluded from the
reliability analysis if it did not load on its hypothesized factor, if it loaded on its
factor but with a value less than .33 and if it had communalities less than .35
coupled with low loading on its factor. Acceptable (moderate) communality
coefficients range between .4 and .7 (Fabrigar et al., 1999) so the above range was
used as a baseline to retain items with communalities of .35 and above. Twenty-
three items (fourteen cognition, six behavior and three affect), rated on a seven-
point Likert scale were used to run the reliability analysis in SPSS. Corrected item
total correlations, squared multiple correlations, average inter item correlations
(IIC), standard deviations, the inter-item correlations matrix, a desirable alpha level
of 0.8 set a priori, and Cronbach‟s alpha if item is deleted informed decisions about
the reliability analysis of the subscales.
Table 5
Summary of reliability analysis for FIMS subscales

Subscale No. Alpha CI95 Avg. SD (IIC) Mean SD

Items IIC
Cognition 14 .92 .90-.94 .46 .14 4.65 .79

Affect 3 .75 .68-.81 .51 .00 4.29 1.39

Behavior 6 .73 .66-.79 .32 .10 4.96 1.03

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 92
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Cognition. This subscale reported a Cronbach‟s alpha of 0.92, (for more

details see table 5). It had moderate to high corrected item total correlations ranging
from .36 to .84, an average inter-item correlation mean of .46 and standard
deviation of .14, in addition to moderate squared multiple correlations (.22 - .76).
Alpha if deleted did not suggest that deleting any existing items would increase
alpha significantly so all fourteen cognition items were retained.

Affect. The affect subscale consisted of three items, with a Cronbach‟s alpha
of 0.75. The corrected item-total correlations were moderate (.57 - .60) and the
same was true for squared multiple correlations that had values ranging from .33 to
.36.
Behavior. The behavior subscale had a reliability coefficient of .73, and
reported moderate item total correlations among its items (.33 -.62). As was the
case with the affect subscale, alpha if deleted did not provide any items that would
increase reliability of the scale if deleted so no items were removed.
All 23 items used to conduct the reliability analysis were retained. Based on
calculations using the Spearman-Brown Prophecy formula, it was determined that
the affect scale would need one additional item to bring its alpha to .8 and behavior
would need to have three additional items to obtain an alpha of .8.

Discussion

So, are teachers biased towards females in mathematics? And if so, to what
extent? The development of FIMS was inspired by two gaps in the literature on
teacher attitudes: the general lack of psychometrically sound instruments to
measure bias as multi-dimensional construct and the limited ability of current
measures to report bias in gradations.

Bias. A comparison of composite scores of FIMS subscales across gender

and grade levels (elementary, middle school and high school) shows that overall
teachers tend to possess neutral attitudes towards female students in their classes.
Items were measured on a 7-point scale with a score of 4 as neutral with values
above 4 indicative of increasing favorable attitudes and vice versa. Table 6 shows
comparisons of means and standard deviations by group. In general, mean scores
for the subscales were within the neutral to slightly favorable range. Behavior
reported slightly higher scores than cognition or affect, indicating that teachers are
more likely to engage in unbiased behaviors in student interactions regardless of
their beliefs or feelings.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 93
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Table 6
Composite sub-scale scores by group compared against FIMS composite score

Elementary Middle High Composite

FIMS Males Females School School School Scores
Sub Scales

Cognition:
4.51 4.72 4.90 4.81 4.50 4.65
Mean 0.71 0.82 0.85 0.83 0.71 0.79

SD
Affect:
4.43 4.23 4.14 4.19 4.38 4.29
Mean 1.08 1.52 1.65 1.42 1.27 1.39

SD
Behavior:
4.77 5.06 4.90 4.83 5.11 4.96
Mean 0.97 1.05 1.06 1.15 0.94 1.03

SD

Results also show little difference in means for cognition and behavior
between male and female teachers or even among teachers in elementary, middle
or high school. Although mean scores for affect were generally lower than for
cognition and behavior (M = 4.29, SD = 1.39 vs. M = 4.65, SD = .79 and M = 4.96,
SD = 1.03, respectively), group means for affect varied more compared to the
group means for cognition and behavior. A comparison of means across gender
reveals higher affect means for male teachers (M = 4.43, SD = 1.08 vs. females M =
4.23, SD = 1.52). Similar comparisons across grade levels also indicate higher affect
means high school teachers (M = 4.38, SD = 1.27) compared to middle school (M
= 4.19, SD = 1.42) and elementary school teachers (M = 4.14, SD = 1.65). These
results suggest that compared to their respective reference groups (gender and
grade level) males (and high school teachers) possess slightly more positive than
neutral feelings towards their students. The differences are not significant though,
since all groups generally were closer to neutral than slightly favorable in their
evaluative judgments, feelings and behaviors towards female students in
mathematics. These results bolster the theory that teachers attitudes are neutral,
and that over time most teachers adjust their expectations accordingly as more
information about the student becomes available (Good, 1987).

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 94
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Bivariate correlations between the subscales (see table 7) yielded non-

significant correlations between cognition and affect and also cognition and
behavior. However, the relationship between affect and behavior was significant.
Additionally, the statistically significant affect-behavior relationship and non-
significant cognition-behavior relationship seems to point to affect, rather than
cognition, as a more likely factor influencing teacher behavior in interactions with
students. That teachers‟ affect toward student engagement and effort influences
teacher behavior towards students suggests that teacher behavior is contextual and is
dependent on student behavior. These results support findings that highlight the
role and influence of student behavior on teacher behavior (Altermatt et al., 1998),
and open the door for researchers to take a closer look at the relationship between
affect and behavior in the classroom.

Table 7
Subscale correlations

Cognition Affect Behavior

Cognition
Pearson Correlation __ -.11 .01
Sig. (2-tailed) .14 .94

Affect
Pearson Correlation __ .31**
Sig. (2-tailed) .00

Behavior
Pearson Correlation __
Sig. (2-tailed) .

Note. ** p <.001 (2-tailed).

Although at first glance the affect-behavior relationship presented in this

study seems to support previous research on these relations, it does not. This is
primarily because of the difference in questions asked and hence conclusions
drawn. Research prior to this study addressed how teachers felt about specific
students and compared that to their behavior whereas this study examined how
teachers felt about specific student behaviors. Past studies reported that students
who teachers liked tended to receive more praise and higher quality questions than
other students (Good & Brophy, 1972; Silberman, 1969), implying teacher
exclusion of students that they lacked positive affect for, from valuable class
discussions or interactions. The moderate, positive, highly significant affect-
behavior relationship in this study on the other hand, demonstrates just the
opposite: that the more strongly teachers feel about lower levels of engagement,
effort and performance in mathematics from a group of students, the more likely
they are to engage in behaviors that are empowering, communicative of high

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 95
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

expectations toward these students, all behaviors which have been previously linked
with students that teachers like (Good & Brophy, 1972) and also with males as
opposed to females (Sadker & Sadker, 1989).
Degree of bias. The FIMS scale is still in its initial stages of development
and as such, cannot offer individual assessment of degree of bias. It does, however,
provide insight on general trends of degree by group. Analyses of means and
standard deviations show that even though most teacher attitudes towards females
in math generally tend to be neutral, there is considerable variability in beliefs,
affect and behavior among teachers. Of notable significance here is the spread in
the measures, especially for the affect subscale.
Standard deviations for the cognition subscale were smaller than those of
affect or behavior, and within close range (SD = 0.71- 0.85) for males, females and
different grade levels. Standard deviations for behavior were equally homogenous
(SD = 0.94 - 1.15) across all groups. On the other hand, the standard deviations for
affect revealed a higher level of heterogeneity among the groups sampled with
values ranging from 1.08 to 1.65 compared to the other subscales. This seems to
suggest that while teachers have similar beliefs about female students there is great
variability with respect to how they feel about the abilities, effort and level of interest
these students show in mathematics, especially among elementary school teachers
(SD = 1.65) and female teachers (SD = 1.52) compared to other grade levels, and
males, respectively.

Limitations and Recommendations for Future Research

The development of affect items for the validation of FIMS was a challenge.
A number of affect items were multidimensional, loading on other factors. This
greatly reduced the number of affect and behavior items that were used in the
reliability analysis. The scale could have benefited from more measures for these
constructs. Additionally, the scale has only undergone an exploratory factor analysis
and is in its early stages of development. Findings presented here cannot and
should not be generalized to the teacher population before confirmatory factor
analyses on different samples have been conducted.
Finally, although FIMS was developed to evaluate teacher bias towards
students in mathematics, it is not mathematics specific. It can and should be
adapted to evaluate teacher beliefs about students belonging to groups in other
disciplines for which negative stereotypes apply, after it has been formalized.

Conclusion

The American Association of University Women‟s latest publication, „Why

so few?‟ reports that despite considerable gains made by women in math and
science, their success in these fields remains obstructed by stereotypes and cultural
biases (AAUW, 2010). The report highlights the role of learning environments on
the interest and achievement of females in math and science. That is, females tend
to perform below their potential when they perceive environmental cues that
highlight negative stereotypes about their ability in math, and that these effects

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 96
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

diminish in low-stereotype contexts where gender similarities in math ability are

emphasized. These results mirror empirical findings that show that situational
factors like sex-composition (Inzlicht & Ben-Zeev, 2003), and features of the
physical environment (Murphy, Steele & Gross, 2007) reduce females‟ sense of
belonging to science, technology, engineering and math (STEM) domains.
The documented impact of learning environments on females‟ achievement
and engagement in mathematics has yielded numerous efforts on the part of
educators, to optimize learning for females. Currently, interventions geared towards
mitigating the effects of stereotypes on female performance in mathematics e.g.
teaching them about how stereotypes negatively affect performance, have been
successful in diminishing these effects (Johns, Schmader & Martens, 2005).
Nevertheless, the achievement of non-threatening mathematics learning
environments cannot be attained through student interventions alone, especially if
teacher bias is still a problem. Because teachers are a significant part of the learning
environment, teacher bias can not only have a detrimental impact on student
achievement and learning outcomes (Brophy, 1987), but it can also affect how
female students are evaluated in these domains (AAUW, 2010). Student-focused
interventions alone are not sufficient because learning environments constitute not
only teachers and students but also interactions between both groups. Therefore,
the effects of these student- focused interventions could further be enhanced by
teacher interventions aimed at alleviating bias. Making teachers aware of their biases
and giving them the tools they need to alleviate these biases will improve the
learning environment for females in mathematics, bringing us closer to creating
mathematics and science friendly learning environments.
However, teacher-based interventions have been scarce, partly because an
evaluative measure of bias that takes into account its multi-dimensional nature has
not been available. As discussed, previous attempts to measure bias ignore the
multidimensional character of attitudes. This study attempted to develop an
instrument that would do exactly that. The evaluation of bias as a multi-dimensional
construct sheds new light on teacher attitudes towards females in mathematics.
Specifically, the use of cognition, affect and behavior as evaluative bases of attitude
provide valuable insight on which component is most likely to drive teacher bias, if
present. Despite the limitations discussed earlier, the Females In Mathematics
Scale, FIMS, is a strong first step in its endeavor to bridge the gap in empirical
findings on teacher bias. As an instrument it has demonstrated relatively strong
psychometric properties for the measures of all 3 dimensions of bias and attempted
to establish degree by demonstrating where teachers lie on a continuum along each
of these dimensions. So far FIMS has not only displayed acceptable levels of
discriminant validity, but its assessment of bias has also lent support to theories by
Good & Brophy (1974), Good (1987), and supported findings by Altermatt et al.,
(1998).
In conclusion, the FIMS has the potential to facilitate robust analysis on
teacher bias in later stages of its development. It holds promise to inform both
theory and practice with respect to the mechanics of attitude formation and change
in regards to teacher bias. Once the validation is completed, the FIMS should be
able to better pin point which dimensions of bias need to be targeted in

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 97
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

interventions for particular groups, as important attitudinal dimensions are included

and measured by the scale.

References

Albarracín, D., Johnson, B. T., & Zanna, M. P. (Eds.) (2005). The handbook of
attitudes. Mahwah, NJ: Lawrence Erlbaum.
Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1993). First grade classroom
behavior: Its short and long-term consequences for school performance.
Child Development, 64, 801-814.
Altermatt, R. E., Jovanovic, J., & Perry, M. (1998). Bias or responsivity? Sex and
achievement-level effects on teachers‟ classroom questioning practices.
Journal of Educational Psychology, 90, 516-527.
Alvidrez, J. & Weinstein, R. S. (1999). Early teacher perceptions and later student
academic achievement. Journal of Educational Psychology, 91, 731-741.
American Association of University Women. (2010). Why so few?: Women in
science, technology, engineering and mathematics. (Research Report. No.
077-10 5M). Washington, DC. Hill, C., Corbett, C., & St. Rose, A.
Antonak, R. F., & Livneh, H. (2000). Measurement of attitudes towards persons
with disabilities. Disability and Rehabilitation, 22, 211–224.
Brophy, J. (1987). Synthesis of research on strategies for motivating students to
learn. Educational Leadership, 45(2), 40-48.
Brophy, J., & Good, T. (1974). Teacher-student relationships: Causes and
consequences. New York: Holt, Rinehart & Winston.
Catsambis, S. (1995). Gender, race, ethnicity and science educations in the middle
grades. Journal of Research in Science Teaching, 32, 243-257.
Cook, G. B., Cameron, L. D., & Tankersley, M. (2007). Inclusive teachers‟
attitudinal ratings of their students with disabilities. The Journal of Special
Education, 40, 230–238.
Cook, D. (1992). Psychological impact of disability. In R. M. Parker & E. M.
Szymanski (Eds.), Rehabilitation counseling basics and beyond (pp. 249–
272). Austin, TX: PRO-ED.
Cornbleth, C., & Korth, W. (1980). Teacher perceptions and teacher-student
interaction in integrated classrooms. Journal of Experimental Education, 48,
259-263.
Denzin, N. K. (1970). The research act in sociology. Chicago: Aldine.
Eccles, J. S. (1993). School and family effects on the ontogeny of children‟s
interests, self perceptions, and activity choice. In J. Jacobs (Ed.), Nebraska
Symposium on Motivation: Vol. 40. Developmental perspectives on
motivation (pp. 145-208). Lincoln, NE: University of Nebraska Press.
Evertson, C., Brophy, J., & Good, T. (1973). Communication of teacher
expectations: Second grade. (Report 92). The University of Texas at Austin,
Austin: Research and Development Center for Teacher Education.
Fabrigar, R. L., Wegener, T.D., MacCallum, C. R., & Strahan, J. E. (1999).
Evaluating the use of exploratory factor analysis in psychological research.
Psychological Methods, 4(3), 272-299.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 98
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

Fang, Z. (1996). A review of research on teacher beliefs and practices. Educational

Research, 38, 47–65.
Findler, L., Vilchinsky, N., &Werner, S. (2007). The Multidimensional Attitudes
Scale Toward Persons with Disabilities (MAS). Rehabilitation Counseling
Bulletin, 50(3), 166-176.
Good, T. L. (1987). Two decades of research on teacher expectations: Research
and future directions. Journal of Teacher Education, 38, 32-47.
Good, T. L., & Brophy, J. E. (1972). Behavioral expression of teacher attitudes.
Journal of Educational Psychology, 63, 617-624.
Hartley, D. (1982). Ethnicity or sex: Teacher definitions of ability and reading
comprehension in an E.P.A. primary school. Research in Education, 28, 9-
24.
Hayton, J. C., Allen, D.G., & Scarpello, V. (2004). Factor retention decisions in
Exploratory Factor Analysis: A tutorial on parallel analysis. Methodological
Resources, 7, 191-205.
Henson, R. (2006). Use of Exploratory Factor Analysis in published research.
Educational and Psychological Measurement, 66(3), 393-416.
Hillman, S. B., & Davenport, G. G. (1978). Teacher-students interactions in
desegregated schools. Journal of Educational Psychology, 70, 545-553.
Hoge, R., & Butcher, R. (1984). Analysis of teacher judgments of pupil
achievement level. Journal of Educational Psychology, 76, 777-781.
Hoge, R. D., & Coladarci, T. (1989). Teacher-based judgments of academic
achievement: A review of literature. Review of Educational Research, 59,
297-313.
Inzlicht, M., & Ben-Zeev, T. (2003). Do high achieving female students
underperform in private? The implications of threatening environments on
intellectual processing. Journal of Educational Psychology, 95(4), 796-805.
Jick, T. D. (1979). Mixing qualitative and quantitative Methods: Triangulation in
action. Administrative Science Quarterly, 24(4), 602-611.
Johns, M., Schmader, T., & Martens, A. (2005). Knowing is half the battle: teaching
stereotype threat as a means of improving women‟s math performance.
Psychological Science, 16 (3), 175-179.
Jones, G. M., & Wheatley, J. (1990). Gender differences in teacher-student
interactions in science classrooms. Journal of Research in Science Teaching,
27, 861-874.
Jussim, L. & Eccles, J. S. (1992). Teacher expectations II: Construction and
reflection of student achievement. Journal of Personality and Social
Psychology, 63, 947-961.
Jussim, L. (1991). Social perception and social reality: A reflection-construction
model. Psychological Review, 93, 54-73.
Ma, A. & Norwich, B. (2007). Triangulation and Theoretical Understanding.
International Journal of Social Research Methodology, 10 (3), 211- 226.
MacCallum, R. C., Widaman, K. F., Zhang, S., & Hong, S. (1999). Sample size in
factor analysis. Psychological Methods, 4, 84-99.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 99
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

McKenzie, J. F., Wood, L. M., Kotecki, E. J., Clark, K. J., & Brey, A. R. (1999).
Establishing content validity: Using qualitative and quantitative Steps.
American Journal of Health and Behavior, 23(4), 311-318.
Murphy, M. C., Steele, C. M., & Gross, J. J. (2007). Signaling threat: How
situational cues affect women in math, science and engineering settings.
Psychological Science, 18(10), 879- 885.
National Science Foundation. (2000). Women, minorities, and persons with
disabilities in science and engineering (NSF Rep. No. 00-327). Arlington,
VA: Bleeker, M. M.
Olson, J. M., & Zanna, M. P. (1993). Attitudes and attitude change. Annual Review
of Psychology, 44, 117-154.
Palardy, J. M. (1969). What teachers believe- what children achieve. Elementary
School Journal, 69, 370-374.
Pett, M. A., Lackey, R. N., & Sullivan, J. J. (2003). Making sense of factor analysis:
The use of factor analysis for instrument development in healthcare
research. London, United Kingdom: Sage.
Rosenfeld, M., & Rosenfeld, S. (2008). Developing effective teacher beliefs about
learners: the role of sensitizing teachers to individual learning differences.
Educational Psychology, 28, 245-272.
Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes,
and other things called emotion: Dissecting the elephant. Journal of
Personality and Social Psychology, 76, 805-819.
Sadker, M., & Sadker, D. (1990). Confronting sexism in the college classroom. In
S. K. Gabriel & I. Smithson (Eds.), Gender in the classroom: power and
pedagogy (pp. 176-187). Urbana: University of Illinois Press.
Silberman, M. L. (1969). Behavioral expression of teachers‟ attitudes toward
elementary school students. Journal of Educational Psychology, 60, 402-
407.
Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd ed.).
Mahwah, NJ: Lawrence Erlbaum.
Tiedemann, J. (2000). Parents‟ gender stereotypes and teachers‟ beliefs as
predictors of children‟s concept of their mathematical ability in elementary
school. Journal of Educational Psychology, 92(1),144-151.
Trouilloud, D., Sarrazin, P., Bressoux, P., & Bois, J. (2006). Relation between
teachers‟ early expectations and students‟ later perceived competence in
physical education classes: Autonomy supportive climate as a moderator.
Journal of Educational Psychology, 98(1), 75-86.
Trouilloud, D., Sarrazin, P., Martinek, T. G., & Guillet, E. (2002). The influence
of teacher expectations on students‟ achievement in physical education
classes: Pygmalion Revisited. European Journal of Social Psychology, 32(5),
591-607.
Velicer, W. F., & Flava, J. L. (1998). Effects of variable and subject sampling as
factor pattern recovery. Psychological Methods, 3, 231-251.

© 2010 Time Taylor Academic Journals ISSN 2094-0734

 100
The International Journal of Educational and Psychological Assessment
August 2010, Vol. 5

About the Authors

Katherine Picho is a doctoral candidate in the Cognition and Instruction program

at the University of Connecticut. Her research interests focus on Stereotype Threat,
and gender differences in leadership style. Katherine‟s areas of interest in
quantitative research methods include hierarchical linear modeling, structural
equation modeling, meta-analysis and psychometrics.
Email: edpsychresearch@gmail.com

Dr. Katrichis is an associate professor in Marketing in the Barney school of

business at the University of Hartford.
Email: Katrichis@hartford.edu

Dr. D. Betsy McCoach is an associate professor in the Measurement, Evaluation

and Assessment program at the University of Connecticut. She has extensive
experience in hierarchical linear modeling, instrument design, factor analysis, and
structural equation modeling. Betsy co-edited Multilevel Modeling of Educational
Data. She has published numerous journal articles and book chapters in the field of
education. Betsy is the current co-editor of the Journal of Advanced Academics.
She also serves on the editorial review boards for the American Educational
Research Journal, the Journal of Educational Psychology, the Journal of
Educational Research, and Gifted Child Quarterly.
Email: betsy.mccoach@uconn.edu

© 2010 Time Taylor Academic Journals ISSN 2094-0734

View publication stats